Otimização em despacho de táxis

Abstract

In the last years several strategies have been developed by companies that intermediate the communication between the passenger and the taxi driver with the call applications. This intermediation is known as assigning a taxi to a passenger and is used in methods of dispatching taxis. To select a taxi it is necessary to adopt some strategy, such as: the taxi nearest to the passenger, the taxi that has the least estimated time arrival, among others. It is worth mentioning that at any moment passengers can cancel the request and taxi drivers can cancel the service. Several factors lead to this type of cancellation, being for the passenger the main reason for the distance of the driver and the driver the destination of the passenger, because the driver only has the knowledge of the destination of the passenger after the accept of the request. In this way, depending on the adopted strategy, we can assign a taxi that does not have the interest in accept this request or canceled. One of the strategies related to the dispatch of the taxi to the passenger is to direct the service requests for the taxis with the greater chance of acceptance. Thus some parameters can be analyzed for decision making regarding the choice of taxi to be assigned to the passenger. In this scenario, the optimization methods can be used to aid the decision-making process and, thus, to select the taxi with the greatest chance for service. Therefore, the objective of this work is to propose an optimization method that uses a mathematical model of linear programming for decision making of which taxi to be assigned to a determining passenger in order to direct the request to the taxi with greater chance of acceptance. In this taxi dispatch method, the following criteria will be adopted: distance between taxi / passenger, taxi's note given by the passenger, taxi driver's cancellation and refusal quantities, taxi's fidelity stamp, go to home and the aging index for the homogenization of taxis / passenger assignments. Actual data were used in the tests and comparisons considering the point of origin of the taxi and the passenger were performed. In the comparisons, the Euclidean distance methods and the actual distance calculation method were used (the latter, carried out through the collaborative mapping project - OpenStreetMap. The analyzes related to the diversity of parameters for taxicab / passenger assignment were performed and demonstrated the influence of the weights assigned to the parameters in the mathematical model and showed that the sending of the requests were to the taxi drivers with greater chance of acceptance.